CoolFace
Modelpublic

FirstPotatoCoder/Genshin_5e_lm

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face
0likes3downloads
config.json123 linesDownload Raw Back to root
1{2  "architectures": [3    "CsmForConditionalGeneration"4  ],5  "attention_bias": false,6  "attention_dropout": 0.0,7  "audio_eos_token_id": 128003,8  "audio_token_id": 128002,9  "bos_token_id": 128000,10  "codebook_eos_token_id": 0,11  "codebook_pad_token_id": 2050,12  "codec_config": {13    "_name_or_path": "kyutai/mimi",14    "architectures": [15      "MimiModel"16    ],17    "attention_bias": false,18    "attention_dropout": 0.0,19    "audio_channels": 1,20    "codebook_dim": 256,21    "codebook_size": 2048,22    "compress": 2,23    "dilation_growth_rate": 2,24    "frame_rate": 12.5,25    "head_dim": 64,26    "hidden_act": "gelu",27    "hidden_size": 512,28    "initializer_range": 0.02,29    "intermediate_size": 2048,30    "kernel_size": 7,31    "last_kernel_size": 3,32    "layer_scale_initial_scale": 0.01,33    "max_position_embeddings": 8000,34    "model_type": "mimi",35    "norm_eps": 1e-05,36    "normalize": false,37    "num_attention_heads": 8,38    "num_filters": 64,39    "num_hidden_layers": 8,40    "num_key_value_heads": 8,41    "num_quantizers": 32,42    "num_residual_layers": 1,43    "num_semantic_quantizers": 1,44    "pad_mode": "constant",45    "residual_kernel_size": 3,46    "rope_theta": 10000.0,47    "sampling_rate": 24000,48    "sliding_window": 250,49    "torch_dtype": "float16",50    "trim_right_ratio": 1.0,51    "upsample_groups": 512,52    "upsampling_ratios": [53      8,54      6,55      5,56      457    ],58    "use_cache": false,59    "use_causal_conv": true,60    "use_conv_shortcut": false,61    "vector_quantization_hidden_dimension": 25662  },63  "depth_decoder_config": {64    "attention_bias": false,65    "attention_dropout": 0.0,66    "backbone_hidden_size": 2048,67    "head_dim": 128,68    "hidden_act": "silu",69    "hidden_size": 1024,70    "initializer_range": 0.02,71    "intermediate_size": 8192,72    "max_position_embeddings": 33,73    "mlp_bias": false,74    "model_type": "csm_depth_decoder_model",75    "num_attention_heads": 8,76    "num_codebooks": 32,77    "num_hidden_layers": 4,78    "num_key_value_heads": 2,79    "rms_norm_eps": 1e-05,80    "rope_scaling": {81      "factor": 32.0,82      "high_freq_factor": 0.0078125,83      "low_freq_factor": 0.001953125,84      "original_max_position_embeddings": 16,85      "rope_type": "llama3"86    },87    "rope_theta": 500000,88    "torch_dtype": "float16",89    "use_cache": true,90    "vocab_size": 205191  },92  "head_dim": 64,93  "hidden_act": "silu",94  "hidden_size": 2048,95  "initializer_range": 0.02,96  "intermediate_size": 8192,97  "max_position_embeddings": 2048,98  "mlp_bias": false,99  "model_type": "csm",100  "num_attention_heads": 32,101  "num_codebooks": 32,102  "num_hidden_layers": 16,103  "num_key_value_heads": 8,104  "pad_token_id": 128256,105  "rms_norm_eps": 1e-05,106  "rope_scaling": {107    "factor": 32.0,108    "high_freq_factor": 0.5,109    "low_freq_factor": 0.125,110    "original_max_position_embeddings": 1024,111    "rope_type": "llama3"112  },113  "rope_theta": 500000,114  "text_vocab_size": 128256,115  "tie_codebooks_embeddings": true,116  "tie_word_embeddings": false,117  "torch_dtype": "float16",118  "transformers_version": "4.52.3",119  "unsloth_version": "2025.7.11",120  "use_cache": true,121  "vocab_size": 2051122}123